Author  Topic 
DrakeStorm
Posts: 1037
Registered: March 2006

Introducing "Tournament" ELO Ranking

Thu, 22 August 2013 05:07


With the help of Sysyphus, I have pulled almost all the tournament data off Tikipedia and put it into an excel spreadsheet.
With this data, we have created an ELO based on just Tournament results  no more over inflated ELOs based on playing sub 1300 players, or mixing of Multi and 2P for the individual maps, etc.
Also the ELO is based on Match wins, not Game wins, since you need to win matches to win tournaments! Currently there are over 5,000 tournament matches with 370 different players logged.
Before I post any results for this TELO, there are a few things still to be done.
1.) If anyone wants to volunteer to get the last data off Tikipedia, please PM me. We are missing NC2005, NC2007, CL2011, CL2012. For some reason those events are logged differently on Tikipedia and I don't see an easy way to cut and paste the data into excel.
2.) I need to determine the "kfactor" for the various events. For those not familiar with ELO, the kfactor is basically the "weight" of the event. For example, should NC be weighted the same as CL or League?
Currently I am using the following numbers, but am open to changing them.
20 = Nation's Cup, Fusion Cup, Champion's League
25 = SPWC, EMC, SMC, AMC, AAT
15 = US League, Euro League
I used NC as the baseline, and decided to weigh the Single Player Events more because they are Best of 7, where Nation's Cup, Fusion, CL were just best of five. And eventhough league is best of 7, I weighed it less because its not as competitive as NC and players might not bring their "A" game every match.
Currently its not in the data, but might be added later, where I can adjust the "kfactor" within an event. By that I mean I can have the Round Robin portion of NC be 20, but the KO phase be higher like 25 (since some years the KO was best of 7). Also could make the QF/SF/F of SPWC higher since that part is best of 9 instead of 7.
Of course there is the option of just making all events the same kfactor. It won't make a huge difference, but it will make some.
3.) What's the point? Mainly to have a better representation of the Top TTR players, and hopefully get more players invloved in tournaments. Also, since the nonUSA maps don't have a lot of data points yet, getting to the top of the TELO ranking for a particular map is much more feasable than the DoW ELO ranking!
[Updated on: Thu, 22 August 2013 10:57]



BAM Vballman20
Posts: 219
Registered: January 2007

Re:Introducing "Tournament" ELO Ranking

Thu, 22 August 2013 16:37


If I understand this correctly, you plan to use this for tournaments. So, that being said, I would not put as much emphasis on nonUS maps unless the tournament involves nonUS maps. Considering NC is the baseline, I would put it and SPWC as the top point values. If you add more value to the nonUS maps, it may give a false point value to someone who is a specialist in a map that is not NC or SPWC related, which is the main map if NC is the baseline. Also, it could even drive people away from playing in maps they are unfamiliar with because they won't want to tank on a map that is unfamiliar to them.
Otherwise, I think it is a great idea! I would recommend just weighting the US based maps higher than the others since NC and SPWC have the most participants and are the flagship tournaments of TTR.
Also, what are you going to do for those players that have never played in a tournament before? How will their initial ELO be calculated?
[Updated on: Thu, 22 August 2013 19:18]



Sysyphus  Pommard
Posts: 3438
Registered: December 2007

Re:Introducing "Tournament" ELO Ranking

Thu, 22 August 2013 20:44


Well, the main point of the "TELO" is to have a particular ranking for each specialty.
There will be a ranking for :
2er USA
2er Euro
2er Swiss
2er Asia
2er AAT
Multi US
Multi Euro
We wish to use each of those for the seeding in each individual tournament.
2er overall TELO can be designed too, but it would be used only for pure information rather than seedings.
But, I'd keep a strong weight on alternative maps for the overall 2er seedings. If you consider yourself to be the best overall player, then go play tournament matches on those other maps!
We could also break the TELO by year, and could replace overall championship in 2er. We could find out who had the better TELO in 2013, 2012 etc..
About the weights :
I think League was, at least in the first seasons, more competitive than CL
How to make TELO more credible ?
We need to fuel the TELO with more matches.
Hence the willingness of revamping league for the US and Euro map.
We could maybe break the year in 2 seasons.
1 season dedicated to pure competition and have the top 16 Telo 2er USA/2er Euro play each other in a league, then TELo 1632 etc...
2nd season in the year could be dedicated to a Swissstyle tournament, where players are paired randomly for each match.
That would meet everybody's wish. Players would meet opponents of their level in season 1, and it would give the possibility for lower ranked players to play strong opponents in season 2.
For Asia, Swiss and AAT. We could have players meeting randomly in the lobby and agree to play a Bo5+ (could be 7/9/23/567) that would be included in the TELO.
[Updated on: Thu, 22 August 2013 20:45]



CIA Truckerteller
Posts: 866
Registered: October 2007

Re:Introducing "Tournament" ELO Ranking

Fri, 23 August 2013 13:33


The idea is great, the implementation is very tricky.
Should playing a lot of games be rewarded ? Using 2005 as a starting point, I can tell you now that Angel will top the list because he started earlier than Schwen for instance. I'm interested in knowing who is the better player based on serious (and recent) results. My guess would be that over the last 34 years in SWPC, NC and League, Schwen was slightly better. Deadin and Gosha might be close.
I asume you start at the beginning and ratings start diverging from there. How do you compare results in 2006 to 2013 ? If I speak for myself, in my first NC and SPWC I didn't even know all the tickets and had maybe 200 games experience on the US map. Does theKid get credit for winning some in the early years, yet it's clear that he and other perceived good players (toutoune) from the earlier years haven't been performing on the same level with the likes of kostas, xbomanx and Tenniskasi in most recent years ?
NC is certainly more relevant than Fusion or CL. Fusion and CL are nice gimmicks, NC is a real tournament. I would say NC 25, CL and Fusion 15. How do you account for 3 EU and 3 US games in Fusion ?
Also, it would be nice if Anu_ would play more. Don't you like us ? Or are you afraid
[Updated on: Fri, 23 August 2013 20:53]



ACP Miguel
Posts: 591
Registered: October 2006

Re:Introducing "Tournament" ELO Ranking

Fri, 23 August 2013 20:13


tricky. but thrilling to see what comes out from that new formula
nice to have some changes



Sysyphus  Pommard
Posts: 3438
Registered: December 2007

Re:Introducing "Tournament" ELO Ranking

Fri, 23 August 2013 20:36


Well, it should be no problem to break the ELO over a particular span of years.
Drake sent me the spreadsheet with the formula, so that I could also do some simulations
Whatever the K factor is, the rankings are pretty consistent.
I checked more particularly the evolution of ELO through time.
(pardon my memory if i'm wrong)
From 20052007, daedin seemed to lead the pack, then Deveric took over until 2010, with daedin, Schwen close in the competition.
Between 2010 and early 2012, there was a fight between chris and Schwen at the top of the rankings.
Before Angel wins SPWC in 2012 he was actually "way behind", something like 100 TELO points. But since then, he's been super hot in tourneys with 2 wins in SPWC and one in League and logically tops current 2er USA TELO at the moment.
All the names you are mentionning (Goscha, Schwen, Angel, daedin with you, Truckerteller) are all close to the top of the rankings, which let me think that it's quite close to our "perceived ranking", and that the rankings are credible.
Xbomanx and Kostas are just around top 10. Kostas might experience a boost with the missing CL results.
About the number of games.
Well, we can not reward either players who play 10 tourney games in 4 years. If you want points, you play tournaments.
In the ranking we have, i didn't see any player benefiting from the fact that they played tons of games (I thinf of the likes of Hecki/Drake or dea).
My concern would be opposite : to see players sitting on their rankings again.
Kostas has not played that many tourneys other than NC, so he's not top 5.
[Updated on: Fri, 23 August 2013 20:43]



DrakeStorm
Posts: 1037
Registered: March 2006

Re:Introducing "Tournament" ELO Ranking

Sat, 24 August 2013 02:35


Truckerteller wrote on Fri, 23 August 2013 04:33  Should playing a lot of games be rewarded ? Using 2005 as a starting point, I can tell you now that Angel will top the list because he started earlier than Schwen for instance.

I think this is incorrect logic. More games doesn't get you a higher ELO. You have to actually win! Also starting later can be a benefit. Say Angel was 1700 ELO when Schwen started playing in tournaments. If both were equal skill, and let's say they each won 1 match against each other, Angel's ELO would drop and Schwen's would rise. Whereas if each started at the same time and had the same ELO neither ELO would change. So Schwen has to play less games to get to top of ELO than Angel did. Sure Schwen has to play in tournaments to get up to Angel's level, but the kfactors are large enough that its not like he would have to play years to do it. 1 win in SPWC would probably do it.
Quote:  I'm interested in knowing who is the better player based on serious (and recent) results.

Which tournaments are "serious" is up for debate, and why I asked for suggestions of different kfactors. The ones I used are how I personally feel about the tournaments that I play in, and my impression from the forums and opponents.
As for who the "current" best player is, if there are enough results, TELO should determine that. I'm not a math expert so I don't know the correct number of data points needed, but for USA 2P we have over 4000. For Asia map there are only 31 results which obviously isn't enough. The only data is from AMC, so whoever won that is going to be on top, but for now that is all we have to go on as to who is the best.
Quote:  I asume you start at the beginning and ratings start diverging from there. How do you compare results in 2006 to 2013 ? If I speak for myself, in my first NC and SPWC I didn't even know all the tickets and had maybe 200 games experience on the US map. Does theKid get credit for winning some in the early years, yet it's clear that he and other perceived good players (toutoune) from the earlier years haven't been performing on the same level with the likes of kostas, xbomanx and Tenniskasi in most recent years.

I think there is a misunderstanding of ELO since there are no "free" points [except at the beginning, but I ran the numbers with a 'tiered' start to TELO points and it made almost no difference]. If thekid was good in 2007, he got the points he deserved. If he is no longer any good in 2012, he will be overrated, lose matches, and his ELO will drop to its proper spot. The only problem with ELO/TELO is getting a high score then sitting on the top without playing. I think for TELO I will do something like DoW does, take them off the published list if they haven't played in awhile (I'm thinking 2 years.. so if haven't played in any tournaments in 2012/2013). They will still have their points and if they play again they will get on the list.
Quote:  NC is certainly more relevant than Fusion or CL. Fusion and CL are nice gimmicks, NC is a real tournament. I would say NC 25, CL and Fusion 15. How do you account for 3 EU and 3 US games in Fusion ?

I didn't play in CL, but I didn't feel like Fusion was a gimmick. I tried to win my match just as much as in other tournaments. In League, I tend not to try as hard, but harder than a random game in the lobby. As for the Fusion 3 EU/3USA match, that one isn't recorded since TELO is by matches, and it doesn't fit in any category. I might break everything down into games at some point to see if there is a difference in TELO and then I would add those games to the appropriate place.
[Updated on: Sat, 24 August 2013 02:42]



DrakeStorm
Posts: 1037
Registered: March 2006

Re:Introducing "Tournament" ELO Ranking

Sat, 24 August 2013 03:16


One other thing I could use help on.
If anyone is familiar with alot of the players over the years, I need to know if anyone has used multiple player IDs in tournaments. It messes up the calculation a bit if they are entered as 2 different people. I found a few people but that is only because they used the same name (Neozen, David, Kotay, and of course Baron!).
And not that it effects the results, but if I have a player listed that isn't their "common" name, I can change it.
496686 ~maf
242671 2Cutter
218794 335d
187991 A_Canadian_eh
252035 acemandope
1016975 Adri 82
194407 Adriano
512894 Alautre
381753 Alleyooper
639035 ameteir
592361 ana49  Chaume
37527 Angel6
121052 Angelo
309474 angil
310082 Annabel
100915 AnTeam
48430 Anu_
175709 aralai
618804 ATLeague
413756 AudreyMuscat
516115 B Flat Minor
271762 Baraka
187160 barbarnaud
505074 barneyg
307615 Baron
245866 Baron #2 (Replaced by Baron 307615)
448514 Baroness
59471 bassie
91262 Bats_
318495 bbandccrock
449693 Bebs3939Meursault
34017 Beka
369933 belelule
330255 Black Widow
108027 blackcoletrain
346936 BlackIsBack
493090 blampy
441955 blubes
361270 Boon le Brack Chardonnay
476678 Brutuss
638565 Bry Guy
47060 BVB 09
160676 calif
543222 Cani
917929 capivng
211073 Cat
365721 cat_6_4_2
144925 CF True
457358 chpemp
105086 chrigu
149121 chrismmm
94836 claro
1277853 CobraZL
783393 Cocco_
357487 Colo
251832 Corflu
342171 Cristiano Ronaldo
567882 Cromze
1079703 cytmike
60145 Da Vince
177058 daedin
487529 dandee
142434 davedavisliégeoise
246283 Davidb9  Mercurey
181459 Davidb9 (Replaced NC2006 to 246283)
145666 dea1
102184 deep_blue
431383 Denmark
317541 Deveric
54315 DieSuperlok
142442 DiLoLAngelus
79305 Diplodokus
197108 dizz
280980 DN
452101 Doc  Pignolo
61366 Donald & Daisy
61068 Donisl
215331 DrakeStorm
222133 Drannoc
122743 Eichi 05
100728 Elric  Sancerre
229963 ElSid
6686 ElSoyokaze
300395 Enix  Poulsard
233968 EPO Moulin à Vent
6 eric
130877 erps
98539 Eurodude
326719 Evelyn3
337966 exceldsr
271300 express
84633 f0uyaya  PicStLoup
1158390 FalkoneZL
222993 Fandan
168229 fczzsc
173885 Federica
225481 Fitzcarraldo
246489 Floutch
334787 foudecoasters
1670 Franck
217371 freddocheOrval
332592 Gaude
522454 GenuineFauxFarm
140312 Geoff
57742 ggfabiChampigny
131190 Giuseppe
495115 gnirlum
134788 Goan
190830 Goscha
545273 Gralsritter
542467 GSV3
146843 Gyuri
243373 HALIK  PRATAVONE
250865 Hauptbahnhof
50603 Hawaienne
609976 haxxli
540407 hbirger
162126 Hecki
155012 Heizer_65
82099 Hermelin
36650 hogika
228775 Hogwarts
165606 hojita
93055 HorstSchlämmer
471752 Hrundi
303324 ilété1foisJurançon
58148 Imotep
565777 ipad
260024 Iron31Patrimonio
110955 Ironhorse
342070 Iseltwald
134081 Ishamael
62915 Itchy
98632 Jac_
406437 jawoi
429639 JenAck
400914 JFGUIGUI graves
394486 JMulcai
511477 joganu
488239 John
436614 JohnM
354297 joni
255509 jonty
39592 José Pastel de Bacalhau
168585 juanfpdm
408636 Julichka
301732 Juuco
142383 Kangoo
441949 Kaogum
158173 Karin1961
555540 KlemZ  Cheval Blanc
597411 Knockando
95742 Koel Player
277410 kolmo
471446 kostasss
56613 Kotay
466897 Kotay #2 (Replaced by Kotay 56613)
1264449 kotjara
189835 KrillStEstephe
654108 Kris_
711961 kristelll
177493 kunibert
87171 ladouce
227959 Lcremita
721826 LeGoupil
548869 Lehmix
218928 Leleline_E
161183 Leleline_J
218597 Leleline_M
425772 Lexx
180027 lirex
168410 lisa171
43732 Loan
1036797 Loca
812967 lollo04
460846 Lucullupus
201953 Magic Hansen
61610 maia03
131262 Make72
47926 Manni
29670 Marilyn
7 MarkoPolo
440255 Marschall Vorwärts
460892 Mary Mary
37613 Masimo
627655 Maskaride  Vouvray
61705 Mason68
260985 Master Yoda
528043 maunaloa
400616 mausmicky2
92040 MaxiJazz
2716 MaximouR
672183 mef23
459263 menno
273078 Miguel
57374 Miriamel
271902 Mnemosyme
383259 Monstarmaster
1212474 Morientes ZL
241099 Mr Bean
30296 msande75
35562 Mudda
220093 myas
13157 mylene
112569 Nanni
97584 Nayeli
452122 near
681060 Néozen  Cairanne
127016 Neozen (Replaced to 681060)
507669 Niamey
49910 nico06
148295 nicole  vin de moselle
661419 Ninou orange
177742 Nomade
247710 Olle Boll
417544 ommie
59094 onyx puffin
51286 Optimus
83506 Origami
361412 pacap  buzet
147198 pacmon
49875 Pammes
231109 PapiXulu
54371 Patrick09
188690 Patterson
255578 paul in motion
215593 pegaso
430806 PenguinPower
108679 Pepetever
83976 Peter de Zeeuw
581993 peterhowitt
697173 Photios
501610 Pifbat69
803941 pingu007
344032 pipodibo2grosplant
341886 pipodibocoteauxdulayon
182870 Pitze
771946 PnavE_81
371260 PopolValDieu
165639 psteinx
341985 pützi
411066 Qorlas
195018 RailBandit
348345 Ramses64
527610 Ravage811
549217 RCec1
244004 RealtorWill
94941 RexCramer
109592 RFAD  SaintEmilion
369208 Rick Rogers
44567 Rick27707
275100 RMarkes
392264 Robin
49588 Robomat
130485 Rocketblizzard
133341 Rogue
99711 Rogue
108073 Rolli
53756 rudi
416461 Rufuz
142164 RV54
224399 SallySue
192149 Saurus
56968 sauvignon57
167193 schnupferl
190627 Schwen
217031 SchwyzerToni
471875 SCM73
160283 Scorps
31980 Sebast
90443 sebbo
87469 shamogi
272694 Shono
234491 SibylleB
51192 Silvertrain
218237 Sir1
132332 sisu
141977 sjaak
185688 Sjoerd de Sloerie
58737 SKMorefield
184511 Snowstar
480170 snuggle
24943 solinsf
384478 Sookie
267105 spartandopeSTW
427025 spugna
478032 Staudenexpress
82309 staycool
149328 stefano98
239384 stemayf
301205 Stephan1972
55437 steve
494896 stocfocs
533053 Stormwarrior
164445 streetcarEndstation
163020 suburu
61402 süszifusz
263823 Suica'sPenguin
160829 Super Chief
1041277 Superciccio
699796 SuperPello
578093 Surfaholic
97085 swissli
406806 Sysyphus
373946 Tablilla
216941 Tazilon
34437 tchols
692064 tchou pinette
134956 Technikerin
298347 Teddy1
103817 TEECiteaux
117146 Tenniskasi
223368 tewe
51410 thadd
517352 The Tinman
556947 THEBEEF
52400 thekid
261076 thesteamwistle
644617 thf5 Muscadet
513363 Thomas  Julienas
48598 ThoMue
217942 Tic
216633 tigévé
329969 Tinkerbell
78264 tiploup  Jasnières
87601 tis11
578192 Toady
259673 Toffer14
177063 Tofra
353316 TOKAR  Pomerol
627633 Toon37 Chinon
85687 toutoune
52391 Tprail
117923 Train
611002 TrainTraindeTchoulesjours
388255 Truckerteller
240526 tufftuff
381180 tumpp1
369884 Tyrana
138056 U 31
51042 U 32
151900 Uhrfriese
290812 Urgestein
89876 UrSpock
472827 uschoen
441209 Val
399610 VanRoyal
394892 Vaz1970
314329 Vballman20
144734 Volleytom
412060 Wagon net  Boudes
169223 wasdenn
286903 wasser
37989 Wernerus
52265 WhiteTrain
119882 Wildfire2099
641584 Will
48051 womble
168151 wostwuist
89816 xavtrax  Saint Amour
451097 xbomanx
340382 XennosSavagnin
346952 xsandrax
134916 yaelka
97850 Yellow
128393 youki
47970 Zaco2
212466 Zeno
299372 Zimo
637334 zinna
438704 zoubaier
469016 Zugle



CIA Truckerteller
Posts: 866
Registered: October 2007

Re:Introducing "Tournament" ELO Ranking

Sat, 24 August 2013 09:43


As you say, with a sufficiently high K factor, things take care of themselves pretty quickly. I was afraid that with a lower k factor, it would be more about the number of winslosses.
I think League A, in the earlier seasons, was just as competitive as SPWC or NC, but it has lost its luster, so overall I would rate it lower than SPWC, NC. Maybe a higher K factor points for League A results.
For me Fusion was just a nice tournament, in fact, I would probably rate my FEUD victory higher than my Fusion victory. I'm almost certain that people brought their A game more in the first few seasons of League A, which unfortunately I didn't play because cheaters were allowed to play league, thanks to spineless TD's.



AAA_dea1
Posts: 2281
Registered: September 2005

Re:Introducing "Tournament" ELO Ranking

Sun, 25 August 2013 16:30


Maybe separate team events from 2 player events?
I care more about being in good shape in FC (CL)  just as much than in NC  because I owe my team the best possible performance.
Other players take their 2p matches more serious.



CIA Truckerteller
Posts: 866
Registered: October 2007

Re:Introducing "Tournament" ELO Ranking

Sun, 25 August 2013 16:37


AAA_dea1 wrote on Sun, 25 August 2013 16:30  Maybe separate team events from 2 player events?
I care more about being in good shape in FC (CL)  just as much than in NC  because I owe my team the best possible performance.
Other players take their 2p matches more serious.

Uhoh, I'll take this as a hint, my captain.
Maybe a different angle: lots of controversy and some cheating incidences in NC, 1 cheating incident and 1 contoversy (final wasn't played) in SPWC. No incidents in FC, CL.


